# Root Cause Analyzer

> Organize root-cause candidates and propose a validation order from symptoms, logs, reproduction steps, and recent changes. Use it when you need to narrow a bug or incident quickly.

- Skill: `lteawoo/root-cause-analyzer` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add lteawoo/root-cause-analyzer`
- Raw SKILL.md: https://api.skillmd.com/api/skills/lteawoo/root-cause-analyzer/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: lteawoo (https://skillmd.com/u/lteawoo)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/lteawoo/root-cause-analyzer

---


# Root Cause Analyzer

## Purpose
Systematically narrow root-cause candidates for a bug or incident and propose the next validation steps.

## Core Principles
1. Do not confuse symptoms with causes.
2. Avoid conclusions without evidence.
3. Present each cause candidate with both likelihood and supporting rationale.
4. Separate reproduction conditions, recent changes, logs, and environment differences.
5. Distinguish immediate mitigation from the long-term fix.
6. Prioritize checks that yield the most information for the least cost.

## Inputs
- Error symptoms or user report
- Reproduction steps
- Logs, error messages, or stack traces
- Recent changes or related diffs
- Environment details such as local, staging, or production

## Workflow
1. Summarize the observed symptoms using confirmed facts only.
2. Separate conditions where the issue reproduces from those where it does not.
3. Inspect recent changes, inputs, state transitions, and external dependency changes.
4. Narrow the list to two to five plausible cause candidates.
5. Document supporting evidence and a falsification method for each candidate.
6. Propose the fastest validation order to reduce uncertainty.
7. If helpful, separate temporary mitigation from the long-term fix direction.

## Output Format

### Symptom Summary
- What has been observed
- Impact scope

### Reproduction Details
- Reproduction steps
- Conditions where it reproduces
- Conditions that remain uncertain

### Cause Candidates
- Candidate 1
  - Supporting evidence
  - How to disprove or confirm it
- Candidate 2
  - Supporting evidence
  - How to disprove or confirm it

### Validation Order
- Logs, code paths, or config to inspect first
- Expected observations

### Temporary Mitigation
- Ways to reduce user impact
- Operational workarounds

### Long-Term Fix Direction
- Areas to modify once the root cause is confirmed
- Tests or regression protection to add

### Remaining Uncertainty
- Additional information needed
- Follow-up questions

## When Information Is Missing
- Keep confirmed facts and hypotheses clearly separated.
- Limit the cause-candidate list to five.
- Limit follow-up questions to three.